Development of a virtual reality system for studying sensorimotor control in mice and characterisation of neural responses to locomotion in the mouse cerebellum

Abstract

The function of the cerebellum has long been known to be essential in motor control and learning. Electrophysiological studies provided evidence of its involvement in locomotion, a behaviour that requires the coordination of multiple muscles and continuous adjustments to adapt to environmental conditions. Over the last forty years, step movements have been shown to be represented at a single cell level during locomotion. Imaging studies revealed that the activity of segregated populations of cerebellar neurons increases during movement as compared to stationary periods. To date, however, little is known as to how cerebellar neural activity modulates with behavioural state, such as locomotion speed. To this end, we recorded activity of cerebellar neurons from locomotion-related areas of the mouse cerebellum and characterised it during virtual reality (VR) behaviour. Experiments were performed using a novel experimental paradigm, combining a VR behavioural task with multi-unit electrophysiological recordings.

As part of the project, a VR system for mice was developed with sufficiently low latency to measure motor behaviour with closed-loop perturbation of sensory information. Detailed protocols were established for the creation and configuration of the system. Behavioural assessment revealed that mice integrated the visual feedback to control their two-dimensional position in the VR environment. In addition, the VR system was integrated with an innovative mechanical perturbation mechanism for generating force fields capable of systematically deviating mouse trajectories in the virtual world. The combination of VR behaviour with a novel perturbation paradigm extends the range of applications of VR systems for studying sensorimotor control.

During VR tasks, recordings from populations of neurons in lobules IV-V of cerebellar vermis were performed. Activity in the majority of the isolated units was found to change with locomotion speed. Some of these units were found to be also modulated with yaw velocity and stride pattern of the mouse. Further analysis revealed that these cells did not constitute separate classes of neurons but, instead, encoded all of these quantities to a greater or lesser extent. Furthermore, locomotion speed could be reconstructed with high accuracy with a simple linear decoder of neural activity of these units.

These findings indicate that cerebellar activity is strongly modulated by kinematic parameters. Further characterisation of the cerebellar neurons via transgenic techniques and novel behavioural tasks (applicable with the VR system presented here) can help increase our understanding of how movement control is achieved through the cerebellum and probe computational theories of sensorimotor integration.